tm1300 NXP Semiconductors, tm1300 Datasheet - Page 210
tm1300
Manufacturer Part Number
tm1300
Description
Tm-1300 Media Processor
Manufacturer
NXP Semiconductors
Datasheet
1.TM1300.pdf
(533 pages)
Available stocks
Company
Part Number
Manufacturer
Quantity
Price
- Current page: 210 of 533
- Download datasheet (7Mb)
TM1300 Data Book
The Y Counter indicates the next pixel from the input
buffer. A new pixel is clocked into the filter registers only
when the Y Counter contents change, which happens
when the Y MSB Counter is loaded with a value greater
than ‘0’. Note that for Y increment values less than 1.0
(up scaling), the change will be caused by carry incre-
ment from the Y LSBs, and a new pixel will not be
clocked into the filter shift register on every Y clock.
For increment values of 2.0 or for values of 1.0 or greater
with carry in (down scaling), multiple new pixels will be
clocked into the filter shift register before the filter inputs
are ready. The number of new bytes needed for the next
pixel is the sum of the Y Increment Integer value and the
carry out of the Y LSB adder. This result is loaded into
the Y MSB Counter. The filter clock is stalled until the in-
puts are ready. The integer value of the increment -- in-
cluding carry -- defines the number of new pixels to be
clocked through the shift register before the filter inputs
are ready for use.
In this discussion, the Y Counter LSBs form a 16-bit bi-
nary number. The upper 5 bits of this 16-bit number form
a 5-bit binary number between 0 and 31 representing a
fractional distance between Y pixels between 0/32 and
32/31. If the new pixel relative distance is 31/32, it is
nearest the right pixel of the two pixels it is between, and
the right 2 pixels will be more heavily weighted than the
left 3.
The horizontal filter shown in
generate a pixel for every integer increment of the Y
Counter. The filter input is always 5 clocks ahead of its
output. The first stage generates the filter term a
using the data from the input block and the a
cient from the coefficient RAM driven by the Y LSBs. The
second stage registers hold the data for X
responding Y LSBs and generate a
stage registers hold the data for X
and generate a
The LSB Register contents can change on every clock.
In the 2:1 scaling example, the LSBs alternated between
0.0 and 0.5. The LSB Counter represents each output
pixel’s x offset value from the input pixel grid. The LSB
Increment value is 16 bits long. The 5 upper bits go to the
coefficient RAMs, and the 11 lower bits provide precision
increment of the LSB Counter for precision in represent-
ing the scaling factor. The 11 lower bits of the LSB Incre-
ment value added to the 11 lower bits of the LSB Counter
determine when to increment the 5 LSBs that drive the
coefficient RAMs and when to clock a new Y pixel into
the filter.
14.5.7.1
For a 5-tap filter, 4 more pixel inputs are needed to the
filter than are generated at the filter output, two before
the first pixel and two after the last pixel. In the worst
case of a window that is exactly N blocks wide and starts
at the first pixel of the first block, two extra blocks must
be read - one at each end of the window - in order to get
these 4 pixels! This is an unavoidable problem with a
multi-tap filter. For an n-tap filter, n-1 extra pixels are
14-12
Loading the extra pixels in the filter
n-2
X
n-2
.
PRODUCT SPECIFICATION
Figure 14-11
n-2
and the X
n+1
X
n+1
is pipelined to
n+1
and its cor-
. The last
n+2
n-2
n+2
coeffi-
LSBs
X
n+2
needed. There are two techniques that avoid this effi-
ciency hit of fetching extra blocks.
1. Move the window edges so they are not within 2 pixels
2. Simulate the edge pixels, such as by mirroring the pair
The ICP uses automatic mirroring to supply these pixels.
Mirroring is used in both horizontal and vertical filter
modes.
14.5.7.2
A line may start and/or end at the edge of the input im-
age. In this case, the two start and/or end pixels needed
for the first and last pixels of the line, respectively, are
missing. The start mirror uses the two pixels to the right
of the first pixel, and the end mirror uses the two pixels to
the right of the last pixel. These pixels are supplied by
controlling the Y counter.
A mirror multiplexer in the 5-tap filter provides mirroring
of one or two pixels at the filter inputs. This mirror multi-
plexer is used for both horizontal and vertical filtering. In
horizontal filtering, the first and last two pixels in the line
are mirrored. The mirror multiplexer is set to the appro-
priate mirror code for the first and last two pixels in the
line. The first two pixels are mirrored for the first two clock
pulses, and the last two pixels are detected using the pix-
el counter for the line.
Mirroring is optional, depending on whether the start or
end of the line is on a window boundary. The DSPCPU
or microprogram must detect this and enable start and/or
end mirroring as required.
14.5.7.3
Figure 14-13
between the SDRAM and the filter for a scaling factor of
1.0. The bus block reads and writes are one fourth of the
filter processing time because the filter processes data at
100 Mpix/sec, and the SDRAM reads and writes blocks
of pixels at 400 Mpix/sec. The SDRAM logic reads the
next block while the current block is being processed.
This also provides the two pixels from the next block re-
quired to finish filtering the current block.
If the scaling factor is greater or less than 1.0. the
SDRAM bus activity will be different. For scaling factors
greater than 1.0, there will be fewer SDRAM reads for
the same number of writes generated by the filter. For
example, a scale factor of 2.0 means that it is necessary
to read only half as many blocks to generate the same
number of output blocks. For a scale factor less than
one, there will be more reads for the same number of
writes. For a scale factor of 0.5, two blocks must be read
for every block of output. If the scale factor is less than 1/
3, more time will be spent reading and writing SDRAM
than filtering.
of a 64 input pixel boundary.
of pixels you have on the other side. This is the only
solution to the problem of starting (or ending) at the
edge of the image, where there are no pixels to the left
(or right) of the image window.
Mirroring pixels at the ends of a line
Horizontal filter SDRAM timing
shows a timing diagram for block data flow
Philips Semiconductors
Related parts for tm1300
Image
Part Number
Description
Manufacturer
Datasheet
Request
R
Part Number:
Description:
Tm Supirbuck Highly Integrated 4a Wide-input Voltage, Synchronous Buck Regulator
Manufacturer:
International Rectifier Corp.
Datasheet:
Part Number:
Description:
NXP Semiconductors designed the LPC2420/2460 microcontroller around a 16-bit/32-bitARM7TDMI-S CPU core with real-time debug interfaces that include both JTAG andembedded trace
Manufacturer:
NXP Semiconductors
Datasheet:
Part Number:
Description:
NXP Semiconductors designed the LPC2458 microcontroller around a 16-bit/32-bitARM7TDMI-S CPU core with real-time debug interfaces that include both JTAG andembedded trace
Manufacturer:
NXP Semiconductors
Datasheet:
Part Number:
Description:
NXP Semiconductors designed the LPC2468 microcontroller around a 16-bit/32-bitARM7TDMI-S CPU core with real-time debug interfaces that include both JTAG andembedded trace
Manufacturer:
NXP Semiconductors
Datasheet:
Part Number:
Description:
NXP Semiconductors designed the LPC2470 microcontroller, powered by theARM7TDMI-S core, to be a highly integrated microcontroller for a wide range ofapplications that require advanced communications and high quality graphic displays
Manufacturer:
NXP Semiconductors
Datasheet:
Part Number:
Description:
NXP Semiconductors designed the LPC2478 microcontroller, powered by theARM7TDMI-S core, to be a highly integrated microcontroller for a wide range ofapplications that require advanced communications and high quality graphic displays
Manufacturer:
NXP Semiconductors
Datasheet:
Part Number:
Description:
The Philips Semiconductors XA (eXtended Architecture) family of 16-bit single-chip microcontrollers is powerful enough to easily handle the requirements of high performance embedded applications, yet inexpensive enough to compete in the market for hi
Manufacturer:
NXP Semiconductors
Datasheet:
Part Number:
Description:
The Philips Semiconductors XA (eXtended Architecture) family of 16-bit single-chip microcontrollers is powerful enough to easily handle the requirements of high performance embedded applications, yet inexpensive enough to compete in the market for hi
Manufacturer:
NXP Semiconductors
Datasheet:
Part Number:
Description:
The XA-S3 device is a member of Philips Semiconductors? XA(eXtended Architecture) family of high performance 16-bitsingle-chip microcontrollers
Manufacturer:
NXP Semiconductors
Datasheet:
Part Number:
Description:
The NXP BlueStreak LH75401/LH75411 family consists of two low-cost 16/32-bit System-on-Chip (SoC) devices
Manufacturer:
NXP Semiconductors
Datasheet:
Part Number:
Description:
The NXP LPC3130/3131 combine an 180 MHz ARM926EJ-S CPU core, high-speed USB2
Manufacturer:
NXP Semiconductors
Datasheet:
Part Number:
Description:
The NXP LPC3141 combine a 270 MHz ARM926EJ-S CPU core, High-speed USB 2
Manufacturer:
NXP Semiconductors
Part Number:
Description:
The NXP LPC3143 combine a 270 MHz ARM926EJ-S CPU core, High-speed USB 2
Manufacturer:
NXP Semiconductors
Part Number:
Description:
The NXP LPC3152 combines an 180 MHz ARM926EJ-S CPU core, High-speed USB 2
Manufacturer:
NXP Semiconductors
Part Number:
Description:
The NXP LPC3154 combines an 180 MHz ARM926EJ-S CPU core, High-speed USB 2
Manufacturer:
NXP Semiconductors